[PR #164] [MERGED] BF16 support #275

New issue

Closed

opened 2026-10-26 21:15:39 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 21:15:39 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepGEMM/pull/164
Author: @RayWang96
Created: 8/15/2025
Status: ✅ Merged
Merged: 8/15/2025
Merged by: @LyricZhao

Base: main ← Head: bf16_support

📝 Commits (1)

9dba5c6 Make various updates and fixes:

📊 Changes

34 files changed (+2293 additions, -495 deletions)

View changed files

📝 README.md (+1 -1)
➕ csrc/apis/gemm.hpp (+471 -0)
➕ csrc/apis/layout.hpp (+85 -0)
➕ csrc/apis/runtime.hpp (+28 -0)
📝 csrc/jit/compiler.hpp (+4 -2)
📝 csrc/jit/handle.hpp (+1 -1)
📝 csrc/jit/kernel_runtime.hpp (+2 -2)
📝 csrc/jit_kernels/heuristics/common.hpp (+3 -2)
📝 csrc/jit_kernels/heuristics/sm100.hpp (+2 -2)
📝 csrc/jit_kernels/heuristics/sm90.hpp (+7 -3)
➕ csrc/jit_kernels/impls/sm100_bf16_gemm.hpp (+143 -0)
📝 csrc/jit_kernels/impls/sm100_fp8_gemm_1d1d.hpp (+3 -2)
📝 csrc/jit_kernels/impls/sm100_fp8_gemm_1d2d.hpp (+3 -2)
➕ csrc/jit_kernels/impls/sm90_bf16_gemm.hpp (+229 -0)
📝 csrc/jit_kernels/impls/sm90_fp8_gemm_1d2d.hpp (+3 -2)
📝 csrc/jit_kernels/impls/smxx_layout.hpp (+55 -8)
📝 csrc/python_api.cpp (+6 -399)
📝 csrc/utils/exception.hpp (+10 -3)
📝 deep_gemm/__init__.py (+38 -10)
📝 deep_gemm/include/deep_gemm/common/scheduler.cuh (+14 -8)

...and 14 more files

📄 Description

Add BF16 support for SM90 and SM100
Refactor Python APIs
Other fixes and code refactoring

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/deepseek-ai/DeepGEMM/pull/164 **Author:** [@RayWang96](https://github.com/RayWang96) **Created:** 8/15/2025 **Status:** ✅ Merged **Merged:** 8/15/2025 **Merged by:** [@LyricZhao](https://github.com/LyricZhao) **Base:** `main` ← **Head:** `bf16_support` --- ### 📝 Commits (1) - [`9dba5c6`](https://github.com/deepseek-ai/DeepGEMM/commit/9dba5c68f79fd879c70751470810d7ec49b25048) Make various updates and fixes: ### 📊 Changes **34 files changed** (+2293 additions, -495 deletions) <details> <summary>View changed files</summary> 📝 `README.md` (+1 -1) ➕ `csrc/apis/gemm.hpp` (+471 -0) ➕ `csrc/apis/layout.hpp` (+85 -0) ➕ `csrc/apis/runtime.hpp` (+28 -0) 📝 `csrc/jit/compiler.hpp` (+4 -2) 📝 `csrc/jit/handle.hpp` (+1 -1) 📝 `csrc/jit/kernel_runtime.hpp` (+2 -2) 📝 `csrc/jit_kernels/heuristics/common.hpp` (+3 -2) 📝 `csrc/jit_kernels/heuristics/sm100.hpp` (+2 -2) 📝 `csrc/jit_kernels/heuristics/sm90.hpp` (+7 -3) ➕ `csrc/jit_kernels/impls/sm100_bf16_gemm.hpp` (+143 -0) 📝 `csrc/jit_kernels/impls/sm100_fp8_gemm_1d1d.hpp` (+3 -2) 📝 `csrc/jit_kernels/impls/sm100_fp8_gemm_1d2d.hpp` (+3 -2) ➕ `csrc/jit_kernels/impls/sm90_bf16_gemm.hpp` (+229 -0) 📝 `csrc/jit_kernels/impls/sm90_fp8_gemm_1d2d.hpp` (+3 -2) 📝 `csrc/jit_kernels/impls/smxx_layout.hpp` (+55 -8) 📝 `csrc/python_api.cpp` (+6 -399) 📝 `csrc/utils/exception.hpp` (+10 -3) 📝 `deep_gemm/__init__.py` (+38 -10) 📝 `deep_gemm/include/deep_gemm/common/scheduler.cuh` (+14 -8) _...and 14 more files_ </details> ### 📄 Description - Add BF16 support for SM90 and SM100 - Refactor Python APIs - Other fixes and code refactoring --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-26 21:15:39 +00:00